Graduation Year


Document Type




Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Computer Science and Engineering

Major Professor

Luther Palmer III, Ph.D.

Co-Major Professor

Yu Sun, Ph.D.

Committee Member

Kyle Reed, Ph.D.

Committee Member

Rajiv Dubey, Ph.D.

Committee Member

Stephen Deban, Ph.D.


Bioinspired Robotics, Motion Planning, Terrain Classification


Legged robots present an incredible opportunity for humanity to conduct dangerous operations such as search and rescue, disaster recovery, and planetary exploration without ever placing themselves in harms way. The ability of a leg to more freely dictate its shape, orientation, and length gives it tremendous mobility and adaptability demanded of a system intended for operation outside of a controlled environment. However, one only need look at the average cat, dog, or friendly neighborhood squirrel to understand the immense gap that exists between what is possible of legged systems and their current set of capabilities.

Areas of study relevant to improving the performance of legged robots outside of the laboratory setting include navigation, path planning, actuator design, gait coordination, sensor development, computer vision, localization and mapping, and so much more. This work chooses to focus on a collection of three inter-related solutions for improving locomotion over uneven terrain: force sensing, workspace analysis and motion planning, and terrain classification. A newly designed, simple force feedback mechanism forms the foundation by providing relevant information about the terrain. To make effective use of the sensor, a method for generating simplified leg workspace representations and an algorithm for robust, offline motion planning are presented. The resulting foot trajectories are suitable for adaptive control in a walking system. The final contribution is a novel approach for terrain classification of terrain height information that gives a hexapod robot the ability to make informed decisions regarding walking parameters. Results from a variety of tests in simulation and hardware prove the effectiveness of the approach and offer a path towards more intelligent control.